ADVANCES IN FLUKA PET TOOLS

Transcription

1 MCMA2017 ADVANCES IN FLUKA PET TOOLS Caterina Cuccagna Tera Foundation (CERN) and University of Geneva Ricardo Santos Augusto, Caterina Cuccagna, Wioletta Kozlowska,Pablo Garcia Ortega, Yassine Toufique, Othmane Bouhali, Alfredo Ferrari, Vasilis Vlachoudis Naples, 18/10/2017

2 Rationale: Why FLUKA for PET FLAIR Complete IDE * for all FLUKA simulation phases (input, geometry editor, debugging, post-processing output visualization) *Integrated Development Environment Physics Models All Hadrons, Leptons On-line evolution of induced radioactivity and dose Benchmarked in the MA energy range (in addition to HEP) See talk G.Battistoni Id. 54 Voxel geometries natively integrated with FLUKA tools for QA MC-TPS DICOM information from clinical CT to FLUKA Voxel geometry Introduction Methods Results Conclusions 2

3 Rationale: Why FLUKA for PET FLUKA code development for (p,d), (n,d) reactions Excitation functions 12 C(p,x) 11 C and 16 O(p,x) 15 O, relevant for PET : Now deuteron formation at low energies is treated directly and no longer through coalescence (Data: CSISRS, NNDC, blue Fluka2011.2, red Fluka2013.0) 11 C version 15 O version 11 C version 15 O version E k (GeV)) E k (GeV)) Introduction Methods Results Conclusions 3

4 Rationale: Why FLUKA for PET Most recent FLUKA code developments Scoring annihilation at rest and activity binning New flag for keeping track for (parent) Isotope: NSS-MIC 2017,Atlanta Introduction Methods Results Conclusions 4

5 Introduction Methods Results Conclusions Rationale: Why FLUKA for PET Most recent FLUKA application for in-beam PET Protons in PMMA M.G. Bisogni INSIDE in-beam positron emission tomography system for particle range monitoring in hadrontherapy, J. Med. Imag. 4(1), (2017), doi: /1.JMI Results on patient presented by E.Fiorina Id. 143

6 Introduction Methods Results Conclusions Integrated in FLAIR Developed in 2013 Tested for conventional PET Generic Radioactive sources Example for small PET scanner Fixed position of the PET scanner Only one image reconstruction algorithm (FBP) FLUKA PET tools : the Origins.. Useful for: Inferring the dose map from the β+ emitter distribution Test new PET design/options P. G. Ortega ANIMMA2013

7 FLUKA PET tools: today Introduction Methods Results Conclusions Rototranslations Integration of post processing and scoring routines in Fluka New PET scanners and validation with NEMA source In-beam PET, beam time structure and acquisition time Studies with RIB (Radioactive Ion Beams) MLEM code

8 WORKFLOW Introduction Methods Results Conclusions 8

9 PET SCANNER MODELS BIOGRAPH, Siemens Introduction Methods Results Conclusions 9

10 Rototranslations Possibility to roto-translate the scanner by defining a translation vector for the center and a rotation vector for the axis Introduction Methods Results Conclusions 10

11 Geometry for New Detectors 10 cm Results on patient presented by E.Fiorina Id cm Introduction Methods Results Conclusions 11

12 WORKFLOW Introduction Methods Results Conclusions 12

13 Specific PET parameters 5 Specific scoring routines FLUKA simulations Output unit Binary or ASCII Energy resolution- Energy window interval around the 511keV (min-max) Acquisition time interval (min-max) [s] Time resolution of the detector [ns] Pulse time of the detector [ns] Hit dead time of the detector [ns] Collection of input parameters Collection of Energy deposited in each crystal Stores info of particle and parents when created. Dumps the buffer into an output file in list mode Implementation of the hit dead time and energy window Introduction Methods Results Conclusions 13

14 WORKFLOW *.dmp Introduction Methods Results Conclusions 14

15 WORKFLOW *.dmp Introduction Methods Results Conclusions 15

16 Coincidences file in list mode The user can perform several analysis : Ex. For in-beam PET with a C12 ion beam In space In time Parent Isotope studies Introduction Methods Results Conclusions 16

17 Coincidences file in list mode The user can perform several analysis : Ex. For in-beam PET with a C12 ion beam In space In time Parent Isotope studies Introduction Methods Results Conclusions 17

18 Coincidences file in list mode The user can perform several analysis : Ex. For in-beam PET with a C12 ion beam In space In time Parent Isotope studies Introduction Methods Results Conclusions 18

19 Coincidences file in list mode The user can perform several analysis on single hit: Ex. For in-beam PET with a C12 ion beam In time In space Parent Isotope studies Introduction Methods Results Conclusions 19

20 Coincidences file in list mode The user can perform several analysis on single hit: Ex. For in-beam PET with a C12 ion beam In time In space Parent Isotope studies Introduction Methods Results Conclusions 20

21 Coincidences file in list mode The user can perform several analysis : Ex. For in-beam PET with a C12 ion beam In space In time O-15 Parent Isotope studies C-11 C-10 B-8 Introduction Methods Results Conclusions 21

22 Coincidences file in list mode The user can perform several analysis : Ex. For in-beam PET with a C12 ion beam In space In time Parent Isotope studies Introduction Methods Results Conclusions 22

23 Coincidences file in list mode The user can perform several analysis on single hit: Ex. For in-beam PET with a C12 ion beam In time In space Parent Isotope studies Introduction Methods Results Conclusions 23

24 WORKFLOW *.dmp Introduction Methods Results Conclusions 24

25 Reconstruction codes FBP (python) Filtered Back Projection Based on the Fourier slice theorem. Simple, fast not accurate enough Available in scikit-image Python package. MLEM Maximum-Likelihood Expectation-Maximization Best estimates the reconstruction image maximizing the likelihood function: Finds the mean number of radioactive disintegrations in the image that can produce the sinogram with the highest likelihood. Iterative, more accurate Integration with STIR Easy to implement Sinogram outputs to STIR STIR Templates are ready for the users, to use different algorithms. Introduction Methods Results Conclusions 25

26 RESULTS 1. Conventional PET for small animals: Example of a commercial scanner (MicroPET P4 scanner) 2. In beam PET in Hadrontherapy with Beta + Radioactive Ion Beams Introduction Methods Results Conclusions 26

27 MicroPET P4 scanner Parameters Crystal dimensions [mm 3 ] P4 scanner 2.2x2.2x10 Detector diameter (cm) 26 Transaxial Field of View (FOV in cm) 18 Axial Field of View (cm) 7.8 Number of detector blocks 168 Total number of detectors (8x8x168) (LSO) ocoincidence - time window: 6 ns ohit dead-time: 500 ns ocoincidence dead-time: 43 ns oenergy window: kev oacquisition time: ns. odetector resolution: 0.14 ns opulse time: 50 ns Introduction Methods Results Conclusions 27

28 MicroPET P4 scanner Voxelized phantom: Digimouse Atlas neuroimage.usc.edu-digimouse Optimization for FLUKA courtesy of M.P.W. Chin of-18 source, generated from USRBIN of Mouse PET image - Introduction Methods Results Conclusions 28

29 MicroPET P4 scanner Voxelized phantom: Digimouse Atlas - Introduction Methods Results Conclusions 29

30 MicroPET P4 scanner orun details: Simulation ran at CERN Cluster. 100 jobs, 5 cycles per job = 500 runs 5 million primaries per run oresults: Average CPU time per cycle: hours ~35 million Coincidences: % trues 0.002% scatters 0% randoms Trues coincidence list file is a 20Gb file... Some hours to process the input files and to reconstruct MLEM up to 70 iterations 30 Introduction Methods Results Conclusions 30

31 MicroPET FOCUS PET Reconstructed images Mouse Phantom CT neuroimage.usc.edu-digimouse FBP (python) Filtered Back Projection MLEM (new code!) Maximum- Likelihood Expectation- Maximization Introduction Methods Results Conclusions 31

32 In-beam PET with RIB Annihilations at rest results:imaging Potential Estimator DOSE ANNIHILATIONS AT REST C-11 C-12 O-15 O-16 SOPB of 1 Gy SOBP in water phantom R. S. Augusto et al.,nss-mic 2016, Strasbourg Introduction Methods Results Conclusions 32

33 In-beam PET with RIB Towards a clinical in-beam PET scenario PET scanner model Siemens Biograph mct as in HIT. R. S. Augusto et al.,ptcog 2017 Yokohama Dose delivery of 1 Gy For SOBPs,11C beam Introduction Methods Results Conclusions 33

34 In-beam PET with RIB Towards a clinical in-beam PET scenario EOB:End of BEAM Introduction Methods Results Conclusions 34

35 In-beam PET with RIB Towards a clinical in-beam PET scenario : offline 25 min Due to the half-life difference between C-11 and O-15 ( 20m & 2m) - C-11 outperforms O-15 in longer acquisitions after irradiation. R. S. Augusto et al.,rad 2017 Introduction Methods Results Conclusions 35

36 In-beam PET with RIB Towards a clinical in-beam PET scenario : online 130 s R. S. Augusto et al.,rad 2017 Introduction Methods Results Conclusions 36

37 In-beam PET with RIB Towards a clinical in-beam PET scenario : in-spill (16 spills) R. S. Augusto et al.,rad 2017 Introduction Methods Results Conclusions 37

38 Conclusions and next steps On going works with PET tools.. Validation of the Clinical Biograph mct Comparison with other codes NEMA Image Quality phantom validation Radioactive Ion beam validation with NIRS experimental results In-beam PET with INSIDE for 12 C and short acquisition time Introduction Methods Results Conclusions 38

39 Thanks for your attention!

40 Back-up slides

41 Conventional PET non invasive imaging modality nuclear medicine field provides three-dimensional (3D) tomographic images of radiotracer distribution within a living subject (molecular imaging) Unstable Parent Nucleus Steps: Proton decays to Neutron positron and neutrino emitted Two antiparallel photons produced Positron combines with e - and annihilates (Line Of Response) Coincidence Unit 1. Radiotracer production 2. Administration of the radiotracer 3. Data Acquisition 4. Image Reconstruction

42 Introduction Methods Results Conclusions Rationale: Why FLUKA for PET FLUKA Monte Carlo code describes b+ emitter distribution for CT-based calculations in patient using Planning CT (segmented into 27 material) and same CT-range calibration curve as TPS (Parodi et al MP 34, 2007, PMB 52, 2007) Experimental cross-sections for b+ emitter production Semi-empirical biological modeling (Parodi et al IJROBP 2007) Convolution with 3D Gaussian kernel (7-7.5 mm FWHM)

43 PROCESSING Michelogram o Arc correction. The radial bin size is corrected for the circular shape of the detector. o Maximum Ring Difference (MRD). The difference between two rings events can be restricted to a maximum value. o Span. Extent of axial data combined. Reduces the size of the stored data. o Mashing factor. Reduction of the angular sampling. Reduces the size of the stored data. o Number of segments. Parameter related to MRD and span number. Defines the number of segments the cells in the Michelogram can be divided.

44 PROCESSING: Coincidences

45 List-Mode ML-EM For certain applications, as when using continuous detectors, where the spatial discretization of the measurements leads to loss of information, it is more appropriate to use a list-mode version of the ML-EM algorithm [*]. Using this method, the main summation runs through the N events in the list-mode data (l = 1,,N). The algorithm is given by: Here, b = 1,, J is the pixel index in the projection operation. The system matrix is the probability that a detected emission from pixel j is detected in the ith detector-pair, corresponding to event l. The list-mode ML-EM algorithm is used for image reconstruction throughout this work. [*] Barrett, H., White, T., Parra, L.: List-mode likelihood. J. Opt. Soc. Am. A 14 (1997) 2914

46 Scoring of PET events During FLUKA simulation: Information of the hits is stored in a buffer and dumped list mode Routines in scoring folder: Usrini.f: Collects the scoring parameters from input. Temporary... Mgdraw.f: Calls petsco.f if energy deposited in PET crystals, and petddt.f and petdmp.f when buffer is full. Temporary... stupre(f)_pet.f: Stores info of particle and parents when created. Petsco.f: Routine to deal with the energy depositions in PET crystals. Petddt.f: Routine that implements the hit dead time and energy window Petdmp.f: Routine that dumps the buffer information in list mode (ascii/bin) (PETCOM): Common with parameters and buffer definitions Udcdrl.f*: Direction biasing. Normally I don't use it, but it is there anyway. Compile: Compile script 46 Introduction Methods Results Conclusions 46

47 Usrini.f (Future PET card) Example of FLUKA card to activate PET routines: Only scoring parameters, no PET geometry involved. If SDUM=SCORE: WHAT(1): Minimum region of PET crystals WHAT(2): Maximum region of PET crystals WHAT(3): Output unit (<0 binary, >0 ascii) WHAT(4): Minimum energy window limit [GeV] WHAT(5): Maximum energy window limit [GeV] If SDUM=SCORE2: WHAT(1): Minimum acquisition time [s] WHAT(2): Maximum acquisition time [s] WHAT(3): Time resolution of the detector [ns] WHAT(4): Pulse time of the detetor [ns] WHAT(5): Hit dead time of the detector [ns] (<0 Paralyzable, >0 Non-paralizable, =0 not used) 47 Introduction Methods Results Conclusions 47